CN105915762A - Noise-pixel adaptive filtering method and noise-pixel adaptive filtering system - Google Patents

Abstract

An embodiment of the invention discloses a noise-pixel adaptive filtering method and a noise-pixel adaptive filtering system. The method comprises the following steps of collecting an image signal of a standard image; converting the image signal of the standard image into a matrix signal; creating a noise interference signal model and calculating a noise signal of the standard image through the noise interference signal model; superposing the matrix signal and the noise signal and taking as a filtering input signal so as to acquire an output result of the filtering system; differentiating the output result of the filtering system and the standard image so as to obtain an error; repeating the above steps, acquiring the plurality of errors, calculating an absolute error of the plurality of errors, and when the absolute error is minimum, acquiring an optimum filtering weight and ending the filtering. By using the technical scheme of the invention, contradiction among a filtering convergence speed, a filtering time-variant tracking ability and a stable state error can be solved; an image can be rapidly and effectively processed; the noise is effectively and rapidly filtered, and simultaneously the stable state error is reduced; main information of the image can be reserved and execution efficiency of an algorithm is increased.

Description

Noise pixel adaptive filter method and noise pixel Avaptive filtering system

Technical field

The present invention relates to Computer Applied Technology field, particularly relate to a kind of noise pixel adaptive filter method and from Adaptive filtering system.

Background technology

In the middle of the Computer Applied Technology field of image procossing, the field that such as television video signal processes, often need One of problem faced by will is the interference for picture quality of various noise (noise) effects.It is known that described noise effect Tackle the interference in picture signal and have the impulsive noise (impulse noise) known to industry, spatial noise (spatial Noise), noise in time domain or other type of noise problem.It is said that in general, when processing these noises, the side generally used Method is depending on using a specific circuit detect and suppress this noise for each noise characteristic.

But, the major defect that existing image filtering method exists has the following aspects: processing speed is fast not, calculation Method execution efficiency is the highest, contradiction in terms of picture noise filter tracking speed and steady-state error, and each above-mentioned defect is existing Image filtering method in can not get well solving always.

In view of this, the embodiment of the present invention is necessary to provide one to filter convergence rate and filter during can solving to take pictures Ripple time-varying follows the tracks of noise pixel adaptive filter method and the noise pixel adaptive-filtering of contradiction between ability and steady-state error System.

Summary of the invention

In order to overcome the defect of above-mentioned background technology, the embodiment of the present invention provides a kind of noise pixel adaptive filter method And Avaptive filtering system, can solve to filter convergence rate and filtering time-varying follows the tracks of the contradiction between ability and steady-state error, And can fast and effeciently process image, while effectively quickly filtering noise, reduce steady-state error, more can retain figure The main information of picture, improves the execution efficiency of algorithm simultaneously.

In order to solve the technical scheme used of the above-mentioned technical problem embodiment of the present invention it is:

A kind of noise pixel adaptive filter method, including:

Gather the picture signal of standard picture；

The picture signal of described standard picture is converted to matrix signal；

Creating noise interferences model, the noise being calculated described standard picture by described noise interferences model is believed Number；

Described matrix signal and described noise signal superposition are tied as filter input signal, the output obtaining filtering system Really；

By poor to the output result of described filtering system and standard picture, obtain error；

Repeat each step above-mentioned and obtain multiple error, calculate the absolute error of the plurality of error, when described absolute mistake When difference is minimum, described terminate this filtering by obtaining optimal filter weights.

Further, described standard picture R represents, shown in described matrix signal such as formula (1):

$R=\left[\begin{array}{cccc}{r}_{11}& r_{12}& \mathrm{...}& r_{1k}\\ r_{21}& r_{22}& \mathrm{...}& r_{2k}\\ .& .& & .\\ .& .& & .\\ .& .& & .\\ r_{g1}& r_{g2}& \mathrm{...}& r_{gk}\end{array}\right]={\[r_{ij}\]}_{g\×k}---\left(1\right)$

In formula (1), g, k are respectively height and the width of described standard picture R；PixelMake C [r_ij] it is with pixel r_ijCentered by the filter window that size is N.

Further, described establishment noise interferences model includes:

(1) sef-adapting filter of second order is used；

(2) W*=[0.8,0.5]^TThe coefficient of wave filter FIR is answered for there being limit for length's unit impact to want；

(3) x (n) be variance be 1, average be 0 Gauss analog input signal；

(4) v (n) be variance be 0.04, average be the white Gaussian noise of 0, and v (n) is uncorrelated with x (n).

Further, the 500th sampled point moment, time-varying system is undergone mutation, and FIR weight vector coefficient becomes W*= [0.4,0.2]^T。

Further, using described matrix signal and described noise signal superposition as filter input signal, obtain filtering system The output result of system, including: using described matrix signal and described noise signal superposition as filter input signal, pass through formula (2), (3) and (4) be calculated the output result of filtering system, described formula (2), (3) and (4) is as follows；

E (n)=d (n)-x^T(n)W(n) (2)

μ (n)=β (1-exp (-α | e (n) |^m)) (3)

W (n+1)=W (n)+2 μ (n) e (n) x (n) (4)

Wherein, described W (n)=[w (n), w (n-1), w (n-2) ..., w (n-L+1)]^TFor wave filter in the power of moment n Weight vector, described X (n) is input stimulus vector, and d (n) is expectation response value, and v (n) is interference noise, and e (n) is error.

Described comparison module is further used for using the plurality of error to calculate described filtering system respectively by formula (5) Weights；

μ (n)=β (1-exp (-α | e (n) |³)) (5)

Wherein, described m is the index coefficient of error e (n), and described m=3 is optimal weight renewing method.

Described method farther includes to reset described filtering system.

A kind of noise pixel Avaptive filtering system, including:

Acquisition module, for gathering the picture signal of standard picture and the picture signal of described standard picture being converted to square Battle array signal；

Processing module, is used for creating noise interferences model, calculates described mark by described noise interferences model The noise signal of quasi-image, for described matrix signal and described noise signal superposition are filtered as filter input signal The output result of system；

Comparison module, for by poor to the output result of described filtering system and standard, obtaining error；And repeat above-mentioned respectively Individual step obtains multiple error, calculates the absolute error of the plurality of error, and when described absolute error minimum, described filtering is tied Bundle.

Further, described standard picture R represents, shown in described matrix signal such as formula (1):

$R=\left[\begin{array}{cccc}{r}_{11}& r_{12}& \mathrm{...}& r_{1k}\\ r_{21}& r_{22}& \mathrm{...}& r_{2k}\\ .& .& & .\\ .& .& & .\\ .& .& & .\\ r_{g1}& r_{g2}& \mathrm{...}& r_{gk}\end{array}\right]={\[r_{ij}\]}_{g\×k}---\left(1\right)$

In formula (1), g, k are respectively height and the width of described standard picture R；PixelMake C [r_ij] For with pixel r_ijCentered by the filter window that size is N.

Further, described processing module establishment noise interferences model includes:

(1) sef-adapting filter of second order is used；

(2) W*=[0.8,0.5]^TThe coefficient of wave filter FIR is answered for there being limit for length's unit impact to want；

(3) x (n) be variance be 1, average be 0 Gauss analog input signal；

(4) v (n) be variance be 0.04, average be the white Gaussian noise of 0, and v (n) is uncorrelated with x (n).

Further, the 500th sampled point moment, time-varying system is undergone mutation, and FIR weight vector coefficient becomes W*= [0.4,0.2]^T。

Further, described matrix signal and described noise signal superposition are believed by described processing module as filtering input Number, obtain the output result of filtering system, including: described matrix signal and described noise signal superposition are made by described processing module For filter input signal, it is calculated the output result of filtering system, described formula (2), (3) by formula (2), (3) and (4) (4) as follows；

E (n)=d (n)-x^T(n)W(n) (2)

μ (n)=β (1-exp (-α | e (n) |^m)) (3)

W (n+1)=W (n)+2 μ (n) e (n) x (n) (4)

Wherein, described W (n)=[w (n), w (n-1), w (n-2) ..., w (n-L+1)]^TFor wave filter in the power of moment n Weight vector, described X (n) is input stimulus vector, and d (n) is expectation response value, and v (n) is interference noise, and e (n) is error noise, L is the exponent number of wave filter.

Described comparison module is further used for using the plurality of error to calculate described filtering system respectively by formula (5) Weights；

μ (n)=β (1-exp (-α | e (n) |³)) (5)

Wherein, described m is the index coefficient of error e (n), and described m=3 is optimal weight renewing method.

Described comparison module is further used for resetting described filtering system.

Having the beneficial effects that of the technical scheme that the embodiment of the present invention is provided: due to the image by gathered standard picture Signal is converted to matrix signal；By creating noise interferences model and calculating the noise signal of described standard picture；By institute State matrix signal and described noise signal superposition as filter input signal, obtain the output result of filtering system；By described filter The output result of wave system system is poor with standard picture, obtains error；Use the weights of filtering system described in described Error Calculation；And And the weights of filtering system when taking described absolute error minimum.Thus, technical scheme disclosed according to embodiments of the present invention, can To solve the contradiction between filtering convergence rate and filtering time-varying tracking ability and steady-state error, and when iterations is identical Time, the steady-state error of technical scheme disclosed according to embodiments of the present invention is minimum；Meanwhile, identical stable state to be obtained by mistake Difference, the iterations of the technical scheme described in the embodiment of the present invention is minimum；Further, when fracture occurs, the present invention is real Execute the technical scheme described in example be the time-varying returning to original steady-state error, i.e. technical solutions according to the invention the soonest with Track ability is the fastest.In a word, the technical scheme described in the embodiment of the present invention, it is possible to fast and effeciently process image, effectively While quickly filtering noise, reducing steady-state error, more can retain the main information of image, that improves algorithm performs effect simultaneously Rate.

Accompanying drawing explanation

Fig. 1 is the flow chart of the present invention a kind of noise pixel adaptive filter method embodiment；

Fig. 2 is the steady-state error comparison diagram of the present invention a kind of noise pixel adaptive filter method embodiment；

Fig. 3 is the convergence rate of the present invention a kind of noise pixel adaptive filter method embodiment and time-varying follow the tracks of ability with Steady-state error comparison diagram；

Fig. 4 is the modular structure schematic diagram of the present invention a kind of noise pixel Avaptive filtering system embodiment；

Fig. 5 is the modular structure schematic diagram of the present invention a kind of noise pixel Avaptive filtering system.

Detailed description of the invention

In order to be illustrated more clearly that the embodiment of the present invention or technical scheme of the prior art, below will comparison accompanying drawing explanation The detailed description of the invention of the present invention.It should be evident that the accompanying drawing in describing below is only some embodiments of the present invention, for From the point of view of those of ordinary skill in the art, on the premise of not paying creative work, it is also possible to obtain other according to these accompanying drawings Accompanying drawing, and obtain other embodiment.

For making simplified form, only schematically show part related to the present invention in each figure, they do not represent It is as the practical structures of product.It addition, so that simplified form readily appreciates, some figure has identical structure or function Parts, only symbolically depict one of them, or have only marked one of them.In this article, " one " not only represents " only this ", it is also possible to represent the situation of " more than one ".

Refer to Fig. 1, one embodiment of the invention provides a kind of noise pixel adaptive filter method.Described a kind of noise Pixel adaptive filter method, comprises the steps:

Step 11, gathers the picture signal of standard picture；The now collection of image can be that the mobile terminals such as mobile phone are taken pictures.

Step 12, is converted to matrix signal by the picture signal of described standard picture；The collection of described standard picture is permissible Taken pictures by mobile terminal of mobile telephone and get, described picture signal is converted to matrix signal and is referred to prior art and carries out Conversion, such as, be converted to the pixel form of matrix signal by the height of the image of described acquisition and width, extracts main pixel and adopts Convert with Matlab software.

Step 13, creates noise interferences model, calculates described standard picture by described noise interferences model Noise signal；Described noise can be the interference of noise pixel during mobile phone photograph.

Step 14, using described matrix signal and described noise signal superposition as filter input signal, obtains filtering system Output result；

Step 15, by poor to the output result of described filtering system and standard picture, obtains error；

Step 16, repeats each step above-mentioned, obtains multiple error, calculate the absolute error of the plurality of error, work as institute When stating absolute error minimum, described by obtain optimal filter weights terminate this filtering.Absolute value when the plurality of error Meansigma methods keep constant time, i.e. can be considered that this error is minimum error.

In the technical scheme that the embodiment of the present invention is provided, owing to the picture signal of gathered standard picture is converted to square Battle array signal；By creating noise interferences model and calculating the noise signal of described standard picture；By described matrix signal and Described noise signal superposition, as filter input signal, obtains the output result of filtering system；Output by described filtering system Result is poor with standard picture, obtains error；Use the weights of filtering system described in described Error Calculation；And take described definitely The weights of filtering system during error minimum.Thus, technical scheme disclosed according to embodiments of the present invention, filtering can be solved and receive Hold back speed and filtering time-varying follows the tracks of the contradiction between ability and steady-state error, and when iterations is identical, according to the present invention The steady-state error of the technical scheme disclosed in embodiment is minimum；Meanwhile, will obtain identical steady-state error value, the present invention is real The iterations executing the technical scheme described in example is minimum；Further, when there is fracture, the technology described in the embodiment of the present invention Scheme be return to the soonest original steady-state error, i.e. technical solutions according to the invention time-varying follow the tracks of ability be the fastest 's.In a word, the technical scheme described in the embodiment of the present invention, it is possible to fast and effeciently process image, is effectively quickly filtering noise While, reduce steady-state error, more can retain the main information of image, improve the execution efficiency of algorithm simultaneously.

Further, described standard picture R represents, shown in described matrix signal such as formula (1):

$R=\left[\begin{array}{cccc}{r}_{11}& r_{12}& \mathrm{...}& r_{1k}\\ r_{21}& r_{22}& \mathrm{...}& r_{2k}\\ .& .& & .\\ .& .& & .\\ .& .& & .\\ r_{g1}& r_{g2}& \mathrm{...}& r_{gk}\end{array}\right]={\[r_{ij}\]}_{g\×k}---\left(1\right)$

In formula (1), g, k are respectively height and the width of described standard picture R；PixelMake C [r_ij] For with pixel r_ijCentered by the filter window that size is N.

Further, described establishment noise interferences model includes:

(1) sef-adapting filter of second order is used；

(2) W*=[0.8,0.5]^TThe coefficient of wave filter FIR is answered for there being limit for length's unit impact to want；

(3) x (n) be variance be 1, average be 0 Gauss analog input signal；

(4) v (n) be variance be 0.04, average be the white Gaussian noise of 0, and v (n) is uncorrelated with x (n).Further, Technical scheme described according to embodiments of the present invention, when the 500th sampled point moment, time-varying system can be undergone mutation, described Limit for length's unit impact is had to want to answer wave filter FIR weight vector coefficient to become W*=[0.4,0.2]^T。

Further, using described matrix signal and described noise signal superposition as filter input signal, obtain filtering system The output result of system, including: using described matrix signal and described noise signal superposition as filter input signal, pass through formula (2), (3) and (4) be calculated the output result of filtering system, described formula (2), (3) and (4) is as follows；

E (n)=d (n)-x^T(n)W(n) (2)

μ (n)=β (1-exp (-α | e (n) |^m)) (3)

W (n+1)=W (n)+2 μ (n) e (n) x (n) (4)

Wherein, described W (n)=[w (n), w (n-1), w (n-2) ..., w (n-L+1)]^TFor wave filter in the power of moment n Weight vector, described X (n) is input stimulus vector, and d (n) is expectation response value, and v (n) is interference noise, and e (n) is error.

Further, described method includes using the plurality of error to calculate described filtering system respectively by formula (5) Weights；

μ (n)=β (1-exp (-α | e (n) |³)) (5)

Wherein, described m is the index coefficient of error e (n), and described m=3 is optimal weight renewing method.According to this Technical scheme described in bright embodiment, the excitation factor error obtained as after the recovery changed as filtering system weights, when absolutely The weights of filtering system acquired when minimizing error make filtering system have the filter effect of optimum.

Described method farther includes to reset described filtering system.When, in filtering, needing upper during filtering every time Filter result once resets, in order to avoid causing the filtering iteration of system to disturb when reusing, the outside of control system is strong simultaneously Reset during interference.Use the clearing scheme described in the embodiment of the present invention can increase the Stability and dependability of system.

Refer to Fig. 2, the stable state of a kind of noise pixel adaptive filter method embodiment described in the embodiment of the present invention is by mistake Difference comparison diagram.The transverse axis of Fig. 2 represents iterations n, and the longitudinal axis represents absolute value Ie (n) I of i.e. noise error e (n) of steady-state error. Three curves in Fig. 2 are respectively the change curve of the value that m value is Ie (n) I when 3,2.75 and 2 from top to bottom.Work as iteration Number of times, and when iterations n value is identical, technical scheme disclosed according to embodiments of the present invention, when m value is 3 Steady-state error Ie (n) I be minimum.

Refer to Fig. 3, the convergence speed of a kind of noise pixel adaptive filter method embodiment described in the embodiment of the present invention Degree and time-varying follow the tracks of ability and steady-state error comparison diagram.The transverse axis of Fig. 3 represents iterations n, and the longitudinal axis represents that steady-state error is i.e. made an uproar Absolute value Ie (n) I of sound error e (n).Three curves 1,2 and 3 in Fig. 3 are respectively the side used described in the embodiment of the present invention The value of Ie (n) I that method, the adaptive filter algorithm (SVSLMS) of variable step regulation rule and the SVSLMS improved are drawn Change curve.From the curve of Fig. 3 it can be seen that to obtain identical steady-state error value, the technical side described in the embodiment of the present invention The iterations of case is minimum；Further, when there is fracture (during n value 500, time-varying system can be undergone mutation), this Technical scheme described in bright embodiment be return to the soonest original steady-state error, i.e. technical solutions according to the invention time It is the fastest for becoming tracking ability.

One embodiment of the invention also provides for a kind of noise pixel Avaptive filtering system, refer to Fig. 4.The present invention one implements A kind of noise pixel Avaptive filtering system that example is provided, including acquisition module 41, processing module 42 and comparison module 43, its In:

Described acquisition module 41, for gathering the picture signal of standard picture and the picture signal of described standard picture being turned It is changed to matrix signal；

Described processing module 42, is used for creating noise interferences model, is calculated by described noise interferences model The noise signal of described standard picture, for described matrix signal and described noise signal superposition are obtained as filter input signal Output result to filtering system；

Described comparison module 43, for by poor to the output result of described filtering system and standard picture, obtaining error；And Repeat each step above-mentioned and obtain multiple error, calculate the absolute error of the plurality of error, when described absolute error minimum, Described filtering terminates.

In the technical scheme that the embodiment of the present invention is provided, owing to the picture signal of gathered standard picture is converted to square Battle array signal；By creating noise interferences model and calculating the noise signal of described standard picture；By described matrix signal and Described noise signal superposition, as filter input signal, obtains the output result of filtering system；Output by described filtering system Result is poor with standard picture, obtains error；Use the weights of filtering system described in described Error Calculation；And take described definitely The weights of filtering system during error minimum.Thus, technical scheme disclosed according to embodiments of the present invention, filtering can be solved and receive Hold back speed and filtering time-varying follows the tracks of the contradiction between ability and steady-state error, and when iterations is identical, according to the present invention The steady-state error of the technical scheme disclosed in embodiment is minimum；Meanwhile, will obtain identical steady-state error value, the present invention is real The iterations executing the technical scheme described in example is minimum；Further, when there is fracture, the technology described in the embodiment of the present invention Scheme be return to the soonest original steady-state error, i.e. technical solutions according to the invention time-varying follow the tracks of ability be the fastest 's.In a word, the technical scheme described in the embodiment of the present invention, it is possible to fast and effeciently process image, is effectively quickly filtering noise While, reduce steady-state error, more can retain the main information of image, improve the execution efficiency of algorithm simultaneously.

Further, described standard picture R represents, shown in described matrix signal such as formula (1):

$R=\left[\begin{array}{cccc}{r}_{11}& r_{12}& \mathrm{...}& r_{1k}\\ r_{21}& r_{22}& \mathrm{...}& r_{2k}\\ .& .& & .\\ .& .& & .\\ .& .& & .\\ r_{g1}& r_{g2}& \mathrm{...}& r_{gk}\end{array}\right]={\[r_{ij}\]}_{g\×k}---\left(1\right)$

In formula (1), g, k are respectively height and the width of described standard picture R；PixelMake C [r_ij] For with pixel r_ijCentered by the filter window that size is N.

Further, described processing module establishment noise interferences model includes:

(1) sef-adapting filter of second order is used；

(2) W*=[0.8,0.5]^TThe coefficient of wave filter FIR is answered for there being limit for length's unit impact to want；

(3) x (n) be variance be 1, average be 0 Gauss analog input signal；

(4) v (n) be variance be 0.04, average be the white Gaussian noise of 0, and v (n) is uncorrelated with x (n).

Further, the 500th sampled point moment, time-varying system is undergone mutation, and FIR weight vector coefficient becomes W*= [0.4,0.2]^T。

Further, described matrix signal and described noise signal superposition are believed by described processing module as filtering input Number, obtain the output result of filtering system, including: described matrix signal and described noise signal superposition are made by described processing module For filter input signal, it is calculated the output result of filtering system, described formula (2), (3) by formula (2), (3) and (4) (4) as follows；

E (n)=d (n)-x^T(n)W(n) (2)

μ (n)=β (1-exp (-α | e (n) |^m)) (3)

W (n+1)=W (n)+2 μ (n) e (n) x (n) (4)

Wherein, described W (n)=[w (n), w (n-1), w (n-2) ..., w (n-L+1)]^TFor wave filter in the power of moment n Weight vector, described X (n) is input stimulus vector, and d (n) is expectation response value, and v (n) is interference noise, and e (n) is error noise, L is the exponent number of wave filter.

Further, described comparison module is used for using the plurality of error to calculate described filtering respectively by formula (5) The weights of system；

μ (n)=β (1-exp (-α | e (n) |³)) (5)

Wherein, described m is the index coefficient of error e (n), and described m=3 is optimal weight renewing method.

Described comparison module is further used for resetting described filtering system.

Technical scheme described in the embodiment of the present invention, mainly according to some picture noises filter that Image Denoising Technology is background Wave method and intelligent algorithm carry out the most effectively processing of digital picture, can be not only used for noise pixel filter during mobile phone photograph Ripple, may be used for the application such as music noise filtering, mobile phone communication noise filtering and simultaneous interpretation noise filtering.

One embodiment of the invention also provides for the structure chart of a kind of noise pixel Avaptive filtering system, refer to Fig. 5.This A kind of noise pixel Avaptive filtering system that a bright embodiment is provided,

Input signal is sequentially input the iteration module of filtering system, obtains the output result of filtering system, and will output Result and reference signal (i.e. noise signal) are made difference and are drawn the excitation factor of filtering method, and this excitation factor changes as weights Principal element, when exporting the result absolute error with reference signal and minimizing, the weights of weights system now reach Excellent, adaptive filter method best results, control module, mainly as peripheral stabilizing control system, mainly realizes answering of system Position, clear operation.

Device embodiment described above is only schematically, and the wherein said unit illustrated as separating component can To be or to may not be physically separate, the parts shown as unit can be or may not be physics list Unit, i.e. may be located at a place, or can also be distributed on multiple NE.Can be selected it according to the actual needs In some or all of module realize the purpose of the present embodiment scheme.Those of ordinary skill in the art are not paying creativeness Work in the case of, be i.e. appreciated that and implement.

Through the above description of the embodiments, those skilled in the art it can be understood that to each embodiment can The mode adding required general hardware platform by software realizes, naturally it is also possible to pass through hardware.Based on such understanding, on State the part that prior art contributes by technical scheme the most in other words to embody with the form of software product, should Computer software product can store in a computer-readable storage medium, such as ROM/RAM, magnetic disc, CD etc., including some fingers Make with so that a computer equipment (can be personal computer, server, or the network equipment etc.) performs each and implements The method described in some part of example or embodiment.

Last it is noted that above example is only in order to illustrate technical scheme, it is not intended to limit；Although With reference to previous embodiment, the present invention is described in detail, it will be understood by those within the art that: it still may be used So that the technical scheme described in foregoing embodiments to be modified, or wherein portion of techniques feature is carried out equivalent； And these amendment or replace, do not make appropriate technical solution essence depart from various embodiments of the present invention technical scheme spirit and Scope.

Claims (14)

1. a noise pixel adaptive filter method, it is characterised in that: including:

Gather the picture signal of standard picture；

The picture signal of described standard picture is converted to matrix signal；

Create noise interferences model, calculated the noise signal of described standard picture by described noise interferences model；

Using described matrix signal and described noise signal superposition as filter input signal, obtain the output result of filtering system；

By poor to the output result of described filtering system and standard picture, obtain error；

Repeat each step above-mentioned, obtain multiple error, calculate the absolute error of the plurality of error, when described absolute error Hour, described by obtain optimal filter weights terminate this filtering.

Method the most according to claim 1, it is characterised in that:

Further, described standard picture R represents, shown in described matrix signal such as formula (1):

$R=\left[\begin{array}{cccc}{r}_{11}& r_{12}& ...& r_{1k}\\ r_{21}& r_{22}& ...& r_{2k}\\ \begin{array}{c}.\\ .\\ .\end{array}& \begin{array}{c}.\\ .\\ .\end{array}& \begin{array}{c}\\ \\ \end{array}& \begin{array}{c}.\\ .\\ .\end{array}\\ r_{g1}& r_{g2}& ...& r_{gk}\end{array}\right]={\[r_{ij}\]}_{g\×k}---\left(1\right)$

In formula (1), g, k are respectively height and the width of described standard picture R；PixelMake C [r_ij] it is With pixel r_ijCentered by the filter window that size is N.

Method the most according to claim 2, it is characterised in that:

Further, described establishment noise interferences model includes:

(1) sef-adapting filter of second order is used；

(2) W*=[0.8,0.5]^TThe coefficient of wave filter FIR is answered for there being limit for length's unit impact to want；

(3) x (n) be variance be 1, average be 0 Gauss analog input signal；

(4) v (n) be variance be 0.04, average be the white Gaussian noise of 0, and v (n) is uncorrelated with x (n).

Method the most according to claim 3, it is characterised in that:

Further, the 500th sampled point moment, time-varying system is undergone mutation, FIR weight vector coefficient become W*=[0.4, 0.2]^T。

Method the most according to claim 3, it is characterised in that:

Further, using described matrix signal and described noise signal superposition as filter input signal, filtering system is obtained Output result, including: using described matrix signal and described noise signal superposition as filter input signal, by formula (2), (3) and (4) are calculated the output result of filtering system, described formula (2), (3) and (4) is as follows；

E (n)=d (n)-x^T(n)W(n) (2)

μ (n)=β (1-exp (-α | e (n) |^m)) (3)

W (n+1)=W (n)+2 μ (n) e (n) x (n) (4)

Wherein, described W (n)=[w (n), w (n-1), w (n-2) ..., w (n-L+1)]^TVow in the weight of moment n for wave filter Amount, described X (n) is input stimulus vector, and d (n) is expectation response value, and v (n) is interference noise, and e (n) is error.

Method the most according to claim 5, it is characterised in that:

Further, described method includes using the plurality of error to calculate the power of described filtering system respectively by formula (5) Value；

μ (n)=β (1-exp (-α | e (n) |³)) (5)

Wherein, described m is the index coefficient of error e (n), and described m=3 is optimal weight renewing method.

Method the most according to claim 5, it is characterised in that:

Described method farther includes to reset described filtering system.

8. a noise pixel Avaptive filtering system, it is characterised in that: including:

Acquisition module, for gathering the picture signal of standard picture and the picture signal of described standard picture being converted to matrix letter Number；

Processing module, is used for creating noise interferences model, calculates described standard drawing by described noise interferences model The noise signal of picture, for described matrix signal and described noise signal superposition are obtained filtering system as filter input signal Output result；

Comparison module, for by poor to the output result of described filtering system and standard picture, obtaining error；And repeat above-mentioned respectively Individual step obtains multiple error, calculates the absolute error of the plurality of error, and when described absolute error minimum, described filtering is tied Bundle.

System the most according to claim 8, it is characterised in that:

Further, described standard picture R represents, shown in described matrix signal such as formula (1):

$R=\left[\begin{array}{cccc}{r}_{11}& r_{12}& ...& r_{1k}\\ r_{21}& r_{22}& ...& r_{2k}\\ \begin{array}{c}.\\ .\\ .\end{array}& \begin{array}{c}.\\ .\\ .\end{array}& \begin{array}{c}\\ \\ \end{array}& \begin{array}{c}.\\ .\\ .\end{array}\\ r_{g1}& r_{g2}& ...& r_{gk}\end{array}\right]={\[r_{ij}\]}_{g\×k}---\left(1\right)$

In formula (1), g, k are respectively height and the width of described standard picture R；PixelMake C [r_ij] be with Pixel r_ijCentered by the filter window that size is N.

System the most according to claim 9, it is characterised in that:

Further, described processing module establishment noise interferences model includes:

(1) sef-adapting filter of second order is used；

(2) W*=[0.8,0.5]^TThe coefficient of wave filter FIR is answered for there being limit for length's unit impact to want；

(3) x (n) be variance be 1, average be 0 Gauss analog input signal；

(4) v (n) be variance be 0.04, average be the white Gaussian noise of 0, and v (n) is uncorrelated with x (n).

11. systems according to claim 10, it is characterised in that:

Further, the 500th sampled point moment, time-varying system is undergone mutation, FIR weight vector coefficient become W*=[0.4, 0.2]^T。

12. systems according to claim 10, it is characterised in that:

Further, described processing module using described matrix signal and described noise signal superposition as filter input signal, To the output result of filtering system, including: described processing module using described matrix signal and described noise signal superposition as filter Ripple input signal, is calculated the output result of filtering system by formula (2), (3) and (4), described formula (2), (3) and (4) as follows；

E (n)=d (n)-x^T(n)W(n) (2)

μ (n)=β (1-exp (-α | e (n) |^m)) (3)

W (n+1)=W (n)+2 μ (n) e (n) x (n) (4)

Wherein, described W (n)=[w (n), w (n-1), w (n-2) ..., w (n-L+1)]^TVow in the weight of moment n for wave filter Amount, described X (n) is input stimulus vector, and d (n) is expectation response value, and v (n) is interference noise, and e (n) is error.

13. systems according to claim 12, it is characterised in that:

Described comparison module is further used for using the plurality of error to calculate the power of described filtering system respectively by formula (5) Value；

μ (n)=β (1-exp (-α | e (n) |³)) (5)

Wherein, described m is the index coefficient of error e (n), and described m=3 is optimal weight renewing method.

14. systems according to claim 13, it is characterised in that:

Described comparison module is further used for resetting described filtering system.